Narrative Values Extractor

Project Overview

The Narrative Values Extractor is an AI-powered tool designed to analyse news articles, op-eds, and other media narratives to extract the underlying value conflicts between different actors. This tool helps users understand the deeper ethical dimensions of public discourse and suggests practical bridging solutions.

This tool demonstrates how AI can be used to make complex value conflicts explicit and actionable, supporting more constructive dialogue and policy development.

Try the Narrative Values Extractor

Key Features

Actor Identification

Automatically identifies and maps all named groups and actors in a narrative, including implicit actors referenced in the text.

Value Extraction

Extracts core values as nouns and noun phrases, distinguishing between stated and inferred values with clear attribution.

Conflict Mapping

Identifies value clashes, narrative devices, and asymmetries of power, voice, risk, or information between actors.

Bridging Solutions

Suggests 2-4 practical, actionable ideas that honour both sides' values without resorting to platitudes.

Example Articles

These articles demonstrate the types of narratives that work particularly well with the Narrative Values Extractor, showcasing clear value conflicts and multiple perspectives.

"Wombat cuddlers" v "wombat shooters" in Taralga

Source: ABC News Background Briefing

Why it works well: This article perfectly illustrates the Narrative Values Extractor's capabilities. It presents clear value conflicts between "wombat cuddlers" (animal welfare advocates) and "wombat shooters" (farmers protecting their livelihoods), with explicit narrative devices like the "cuddlers vs shooters" framing. The story maps multiple actors with distinct values, legitimacy claims, and red lines.

Key value conflicts: Animal welfare ↔ Property rights, Conservation ↔ Economic survival, Urban values ↔ Rural traditions

Read Article Try Analysis

Technical Implementation

How Version 1 Works

The Narrative Values Extractor Version 1 is a proof-of-concept AI tool built using ChatGPT's custom GPT functionality. Here's how it operates:

Input Processing

• Accepts website URLs or pasted text content
• Handles paywalled content by requesting pasted text
• Captures title and date information when available

Analysis Process

• Enumerates all named groups and actors
• Extracts values as nouns/noun phrases
• Distinguishes stated vs inferred values

Evidence Discipline

• Prefers quotations and clear facts from sources
• Provides inline citations when browsing is enabled
• Avoids over-citation while maintaining accuracy

Output Generation

• Creates structured markdown reports
• Maps value clashes and narrative devices
• Suggests concrete bridging solutions

Version 1 Limitations

This is a proof-of-concept demonstration created to show how AI might help extract and map value conflicts in media narratives. It's designed for analytical purposes and should be treated as a tool for understanding discourse rather than definitive analysis.

Status: Proof of Concept Version: 1.0